Quantifying Photorealism in Simulated Data with Gans

1. A system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed to: receive a synthetic image at a first deep neural network; and determine, via the first deep neural network, a prediction indicative of whether the synthetic image is machine-generated or is sourced from a real data distribution; wherein the prediction is used as feedback to a generator included in the first deep neural network to train the generator and then communicated from the generator to a discriminator included in the first deep neural network to update the discriminator training; and wherein the prediction comprises a quantitative measure of photorealism of synthetic image.

2. The system of claim 1 , wherein the quantitative measure of the photorealism indicates how close the synthetic image corresponds to an image from the real data distribution.

3. The system of claim 1 , wherein the synthetic image depicts a plurality of objects.

4. The system of claim 3 , wherein the synthetic image depicts the plurality of objects corresponding to an image view of a simulated image.

5. The system of claim 4 , wherein the simulated image is generated by an image generator.

6. The system of claim 5 , wherein the image generator comprises a gaming engine.

7. The system of claim 4 , wherein the synthetic image is generated via a second deep neural network based on the simulated image.

8. The system of claim 7 , wherein the second deep neural network comprises an encoder-decoder neural network and the first deep neural network comprises a convolutional neural network.

9. The system of claim 1 , wherein the processor is further programmed to: determine that the synthetic image is sourced from the real data distribution when the prediction is greater than or equal to a predetermined threshold.

10. The system of claim 9 , wherein the processor is further programmed to: store the synthetic image in a database when the prediction is greater than or equal to the predetermined threshold.

11. A method comprising: receiving a synthetic image at a first deep neural network; and determining, via the first deep neural network, a prediction indicative of whether the synthetic image is machine-generated or is sourced from a real data distribution; wherein the prediction is used as feedback to a generator included in the first deep neural network to train the generator and then communicated from the generator to a discriminator included in the first deep neural network to update the discriminator training; and wherein the prediction comprises a quantitative measure of photorealism of synthetic image.

12. The method of claim 11 , wherein the quantitative measure of the photorealism indicates how close the synthetic image corresponds to an image from the real data distribution.

13. The method of claim 11 , wherein the synthetic image depicts a plurality of objects.

14. The method of claim 13 , wherein the synthetic image depicts the plurality of objects corresponding to an image view of a simulated image.

15. The method of claim 14 , wherein the simulated image is generated by an image generator.

16. The method of claim 15 , wherein the image generator comprises a gaming engine.

17. The method of claim 13 , wherein the synthetic image is generated via a second deep neural network based on the simulated image.

18. The method of claim 17 , wherein the second deep neural network comprises an encoder-decoder neural network and the first deep neural network comprises a convolutional neural network.

19. The method of claim 18 , further comprising: determining that the synthetic image is sourced from the real data distribution when the prediction is greater than or equal to a predetermined threshold.

20. The method of claim 19 , further comprising: storing the synthetic image in a database when the prediction is greater than or equal to the predetermined threshold.